Working Paper SerieS. Is there a role for domestic demand pressure on export performance? NO 1594 / september 2013

Transcription

1 Working Paper SerieS NO 1594 / september 2013 Is there a role for domestic demand pressure on export performance? Paulo Soares Esteves and António Rua the Competitiveness Research Network In 2013 all ECB publications feature a motif taken from the 5 banknote. NOTE: This Working Paper should not be reported as representing the views of the European Central Bank (ECB). The views expressed are those of the authors and do not necessarily reflect those of the ECB.

2 CompNet The Competitiveness Research Network This paper presents research conducted within the Competitiveness Research Network (CompNet). The network is composed of economists from the European System of Central Banks (ESCB) - i.e. the 27 national central banks of the European Union (EU) and the European Central Bank a number of international organisations (World Bank, OECD, EU Commission) universities and think-tanks, as well as a number of non-european Central Banks (Argentina and Peru) and organisations (US International Trade Commission). The objective of CompNet is to develop a more consistent analytical framework for assessing competitiveness, one which allows for a better correspondence between determinants and outcomes. The research is carried out in three workstreams: 1) Aggregate Measures of Competitiveness; 2) Firm Level; 3) Global Value Chains CompNet is chaired by Filippo di Mauro (ECB). The three workstreams are headed respectively by Chiara Osbat (ECB), Antoine Berthou (Banque de France) and João Amador (Banco de Portugal). Julia Fritz (ECB) is responsible for the CompNet Secretariat. The refereeing process of this paper has been coordinated by Editorial Board of the ECB Working Paper Series, led by Philipp Hartmann. The paper is released in order to make the research of CompNet generally available, in preliminary form, to encourage comments and suggestions prior to final publication. The views expressed in the paper are the ones of the author(s) and do not necessarily reflect those of the ECB, the ESCB, and of other organisations associated with the Network. Acknowledgements The authors would like to thank Carlos Robalo Marques, Luca Opromolla and Maximiano Pinheiro for comments and suggestions. The authors would also like to thank participants in the Competitiveness Research Network meeting held at the European Central Bank in December 2012 and Banco de Portugal seminar, where a preliminary version of the paper was presented. The analysis, opinions and findings of this paper represent the views of the authors, which are not necessarily those of the Banco de Portugal. Paulo Soares Esteves Banco de Portugal; pmesteves@bportugal.pt António Rua Banco de Portugal; antonio.rua@bportugal.pt European Central Bank, 2013 Address Kaiserstrasse 29, Frankfurt am Main, Germany Postal address Postfach , Frankfurt am Main, Germany Telephone Internet Fax All rights reserved. ISSN EU Catalogue No (online) QB-AR EN-N (online) Any reproduction, publication and reprint in the form of a different publication, whether printed or produced electronically, in whole or in part, is permitted only with the explicit written authorisation of the ECB or the authors. This paper can be downloaded without charge from or from the Social Science Research Network electronic library at Information on all of the papers published in the ECB Working Paper Series can be found on the ECB s website, europa.eu/pub/scientific/wps/date/html/index.en.html

3 Abstract Traditionally, exports behavior is modeled only as a function of the foreign demand and the real exchange rate. However, it is by now widely acknowledged that these variables are not able to fully explain exports developments. This paper suggests considering domestic demand pressure as an additional variable, revisiting its economic rationale and assessing its empirical importance. In particular, we consider the Portuguese case and find that domestic demand developments are relevant for the short-run dynamics of exports. Moreover, it is found that this relationship is asymmetric, being stronger and more significant when domestic demand is falling than when it is increasing. Keywords: Exports; Domestic Demand Pressure; Error Correction Models; Asymmetry. JEL classification: C22, C50, F10. 1

4 Non-Technical Summary Traditionally, exports behavior is modeled only as a function of the foreign demand and the real exchange rate, i.e. considering demand factors. However, it is by now widely acknowledged that traditional variables are not able to fully explain exports developments. Assuming some substitutability between foreign and domestic sales, this paper suggests domestic demand pressure as an additional explanatory variable. Domestic conditions can influence firms willingness or ability to supply exports and we review the economic reasoning that may underlie a negative relationship between domestic demand behavior and exports. In periods of slacking domestic demand firms may try to compensate for the decline in domestic sales through increased efforts to export, while in boom periods production can be mainly sold on the domestic market. Furthermore, we also discuss the potential asymmetry of this relationship. For instance, when domestic demand increases firms may not leave foreign markets because they already supported some sunk costs. Thus, when modeling export performance, one should take into account not only the driving forces of external demand but also domestic demand, as the former affect exports from the demand side and the latter from the supply side. Presently, this effect can be particularly important given the strong decline of domestic demand in some European countries under a major economic adjustment process. Besides the traditional positive relationship between imports and domestic demand, this effect reinforces the role of domestic demand on external imbalances adjustment. We address the Portuguese case and we find that domestic demand behavior is relevant for modeling the short-run dynamics of exports. In particular, the estimation results suggest that lagged domestic demand developments affect significantly and negatively export performance. Moreover, we find that this relationship is asymmetric, being stronger when domestic demand is falling than when it is increasing. All these findings are supported by a thorough sensitivity analysis. 2

5 1 Introduction Typically, export performance is modeled as a function of the foreign demand for a country s output and a country s price competitiveness indicator. In general, the foreign demand is proxied by the evolution of imports in the trade partners and its relative evolution vis-à-vis exports is used as a measure of market share developments. The relative price advantage of a country over its competitors is often captured by the real exchange rate. Ceteris paribus, a depreciation makes the country s products cheaper relative to its competitors in the foreign market, which will raise the corresponding demand and increase exports leading to an increase of the market share. These factors are essentially related to the demand side. In fact, most studies do not consider supply side variables explicitly when modeling exports. However, it has been recently widely acknowledged that such determinants are far from able to fully explain export performance (see, for example, Fagan et al. (2001, 2005), di Mauro and Forster (2008), European Commission (2010), Dieppe et al. (2012)). Such evidence reinforces the need to search for other factors that may influence exports dynamics. In line with some previous literature, this paper suggests considering domestic demand pressure as an additional explanatory variable. In fact, it is likely that domestic conditions influence firms willingness or ability to supply exports. In a context of high domestic demand pressure, firms will work at full capacity and will not be able to follow, in the short-run, external demand increases. In contrast, during a domestic recession, firms will be able to allocate more resources to exports. In other words, in periods of slacking domestic demand firms try to compensate for the decline in domestic sales through increased efforts to export while in boom periods production can be mainly sold on the domestic market. Early work focusing on the short-run effects of domestic demand pressure on exports includes Ball et al. (1966), Smyth (1968), Artus (1970, 1973), Zilberfarb (1980), Faini (1994), Sharma (2003), among others. In those studies it was found a significant negative ef- 3

6 fect of domestic demand pressure on exports for several countries, including the United Kingdom, the United States, Germany, Israel, Turkey, Morocco and India. Thus, when modeling export performance, one should take into account not only the driving forces of external demand but also domestic demand, as the former affect exports from the demand side and the latter from the supply side. More recently, there has been theoretical and empirical research at the firm level that allows for a better understanding of the negative relationship between domestic demand and exports. Such developments will also contribute to influence the macroeconometric modeling of exports. In this paper, we revisit the theoretical role of domestic demand pressure on exports and assess its importance on modeling the export performance of the Portuguese economy. 1 Besides the recent literature at firm level, such assessment is also motivated by the fact that the standard exports modeling approach is unable to capture properly the Portuguese export performance over the most recent period. In particular, it has been observed a significant and continuous increase of exports market share which cannot be explained by developments on price competitiveness indicators. Such phenomenon is happening along with a dramatic fall of domestic demand. In fact, this relationship could be particularly important in the current economic situation, not only in Portugal but also in other European countries under macroeconomic adjustment and facing strong declines of domestic demand. Following a macroeconometric approach, it is found that domestic demand behavior is relevant for modeling the short-run dynamics of Portuguese exports. In particular, the estimation results suggest that lagged domestic demand developments affect significantly and negatively export performance. Moreover, it is found that such relationship is asymmetric, being stronger when domestic demand is falling than when it is increasing. 1 An earlier attempt to assess the role of domestic demand pressure on Portuguese exports behaviour can be found in Cartaxo (1985). 4

7 All these findings are supported by a thorough sensitivity analysis. The paper is organized as follows. In section 2, we revisit the theoretical rationale for the role of domestic demand on export performance. The empirical results for the Portuguese case are discussed in section 3. Finally, section 4 concludes. 2 The rationale for the role of domestic demand pressure There are several intuitive macroeconomic arguments that can explain a negative relationship between domestic demand and exports. One possible reason is related to the demand side. In particular, when domestic demand is growing, the associated inflationary pressures can lead to a decline of the price competitiveness of exports. 2 There are also arguments from the supply side. On the one hand, during the business cycle, the availability of resources for the exporting sector is affected, which can influence the export performance. On the other hand, in the presence of very different developments for domestic and foreign markets, investment will be most probably oriented to activities that draw more heavily on the most dynamic market. As pointed out in earlier work by, for example, Ball (1961) and Artus (1970), an analysis at the firm level is useful for the identification of the main factors underlying the relationship between domestic demand and exports at the macroeconomic level. Such a microeconomic approach makes it possible consider factors not usually taken into account at the macroeconomic level. In particular, assuming that firms are not price-takers in the 2 In practice, this argument can be potentially refuted since one can argue that this effect is already taken into account through the real exchange rate. However, one can also argue that prices are relatively rigid in the short run (especially downward) so that they do not reflect adequately changes in domestic demand pressure (as pointed out, for example, by Zilberfarb (1980)). 5

8 domestic market can allow one to provide an economic reasoning to the negative relationship between domestic demand and exports. Let us consider the plausible case where export sales are less profitable (because of transport costs, for example) or require greater effort (such as advertising and marketing) than domestic sales. Since capacity production is limited, in the short run, firms will tend to prefer selling to the domestic market if domestic demand increases in detriment of export sales. In contrast, if domestic demand falls firms will try to increase exports as the costs of excess capacity may outweigh the additional costs and effort of selling in the foreign market rather than at home. Additionally, Ball (1961) presents a simple model in which a short-run profit maximizing firm is faced with negatively sloped demand curves in domestic and foreign markets. In this model, where firms have some degree of pricing power in both markets, firms set marginal revenue from exports equal to marginal revenue from domestic sales and marginal costs. In such a framework, given a domestic demand change, the direction of the change in exports volume will depend on the slope of the firm s marginal cost curve in the neighborhood of the initial equilibrium output and the magnitude of this change is determined by the slope of the marginal cost curve and by the demand elasticities in the domestic and foreign markets. Consider, for instance, the case where domestic demand increases. If the marginal costs rises (falls) as output increases, exports will decrease (increase). The steeper is the slope of the marginal cost curve and the higher the foreign demand elasticity, 3 the larger will be the change in exports. In contrast, the higher is domestic demand elasticity, the smaller will be the change in exports. In practice, Ball (1961) suggests that for most British exporting firms, the slope of the marginal cost curve around the initial equilibrium output and the demand elasticities in both markets were such that, under the above 3 Note that, under the usual small open economy assumption, the foreign demand is perfectly elastic and therefore the effect is likely to be high. 6

9 mentioned model, a decrease in domestic demand would lead to an increase of exports. Some of the arguments raised several years ago seem to be reemerging with the recent theoretical and empirical research at the firm level. A key feature of the standard models of international trade is the assumption of constant marginal costs, which allows domestic and foreign markets to be treated independently. However, Vannoorenberghe (2012) provides strong supporting evidence that sales in the domestic and export market are negatively correlated using a large panel of French firms. This empirical pattern can be explained using a Melitz (2003) type of model of international trade with demand uncertainty in which firms face market-specific shocks and short-run convex costs of production. In such a framework, it is found that firms react to a shock in one market by adjusting their sales in the other market. As stressed by Vannoorenberghe (2012), such a result casts doubts on the standard hypothesis that firms face constant marginal costs and maximize profits on the domestic and export markets independently of each other. In fact, there is an emerging literature that emphasizes the fact that the presence of capacity constraints or increasing marginal costs may generate a negative correlation between domestic and export sales (see, for instance, Ahn and McQuoid (2012) for a thorough discussion of the sources of export-domestic sales trade-offs). In this respect, Blum et al. (2011) found such a negative link with Chilean firm level data, Soderbery (2011) reports similar empirical evidence for Thailand whereas Ahn and McQuoid (2012) found a similar pattern for the Indonesian firms. This suggests that, in face of a negative domestic demand shock, existing firms would sell relatively less to the domestic market and more to foreign markets. 4 Furthermore, it seems plausible to believe that new investment by existing firms or new 4 One should note that, in the Portuguese case, there seems to be scope for this relocation in terms of market destination. For instance, in the manufacturing sector, in 2010, only one third of the firms were exporting and for those firms the export intensity, defined as the exports to sales ratio, was on average around 30 per cent. 7

10 firms entering the market would tend to be export oriented given the depressed domestic demand conditions, strengthning the negative relationship between domestic demand and exports. 5 In addition, the relationship between domestic demand and exports performance may be asymmetric. Consider the case where firms need to pay a sunk cost to enter a foreign market (as in Baldwin and Krugman (1989)). For instance, Roberts and Tybout (1997) have found that sunk entry costs are significant using data on Colombian plants. In the presence of sunk costs and uncertainty, the decision to start or stop exporting can be studied following the literature on investment under uncertainty. Based on Dixit (1989) model, Impullitti et al. (2012) consider the export market entry and exit decision in a general equilibrium framework with heterogeneous firms. One can argue that, in the presence of a negative domestic demand shock, itmaybeworthwhileforfirms to pay the sunk entry costs and start exporting. However, in order to avoid repaying the entry cost, incumbent exporters may not leave easily the export market if economic conditions turn to be less favorable. In fact, there is empirical evidence supporting the idea of a noteworthy persistence in a firm s export status (see, for example, Bernard and Wagner (2001) for German firms, Campa (2004) for Spanish firms and Bernard and Jensen (2004) for U.S. plants). 5 Nevertheless, one should bear in mind that, both at the theoretical and empirical levels, the relationship between exports and domestic sales is not clear cut. For instance, on the theoretical front, one may have a positive correlation between domestic sales and exports through overall efficiency improvements (as in the case of learning-by-doing or learning-by-exporting effects). Another reason that may induce such a positive link is related to liquidity constraints (see Berman et al. (2011)). Concerning the empirical front, the results by Berman et al. (2011) suggest that exports and domestic sales are complementary for a panel of French firms. 8

11 3 The Portuguese experience 3.1 Data Due to data availability constraints the sample period herein studied ranges from the first quarter of 1980 up to the second quarter of 2012, which corresponds to 130 quarterly observations. The foreign demand index for Portugal, computed as a weighted average of the import volumes of the Portuguese trade partners, is provided by the European Central Bank and it has been adjusted for the impact of the tax fraud in the United Kingdom (see, for example, Bank of England (2006)). The series for exports and domestic demand, in real terms, correspond to the release of Quarterly National Accounts by INE in September 2012, covering the period since the first quarter of 1995 up to the second quarter of 2012, which have been extended with the historical quarterly series available at the Banco de Portugal website. The real effective exchange rate for Portugal is based on GDP deflators and is provided by the ECB under the branch Harmonized Competitiveness Indicators at the ECB Statistical Data Warehouse. Since it is available only since the first quarter of 1993, we considered a proxy for the previous period (based on the major Portuguese trade partners) to obtain a longer time series. An increase of the real effective exchange rate corresponds to a real appreciation. 3.2 Evolution of the Portuguese exports market share Figure 1 presents the evolution of the Portuguese exports market share from 1980 onwards. During the 80 s, and in particular after 1982, Portuguese exports volume grew almost twice the foreign demand. The index measuring export market shares (1999Q1 = 100) increased from a minimum close to 55 at the beginning of 1982 to a figure close to 105 in After some decline in 1991 and 1992, exports market share returned to an upward trend, reaching its historical maximum at the end of During the second half of the 9

12 90 s, there was a continuous decline of the market share. Concerning its evolution in the current century, it should be mentioned that exports market share stood relatively stable until mid-2004, which was followed by a huge decline in a context of the phasing-out in 2005 of the trade barriers established in the Multifiber Agreement on Textiles and Clothing 6 and reflecting the impact related to the entrance of China in the World Trade Organization (WTO) in After this adjustment, exports market share remained once again relatively stable until the end of Thereafter, exports performance improved significantly, with market share increasing consecutively over several quarters in an amount of almost 10 per cent. A natural question that arises is how can one explain this recent increase of market share considering the usual export modeling strategies. Firstly, it is important to question if this increase of exports market share is or is not statistically relevant. The answer seems to be affirmative. Despite the volatility of the quarterly series, this recent increase appears to be noteworthy. The observed increase during five quarters in a row since 2011Q1 is the third longest period of consecutive gains in exports market share. Until 1985Q2 exports market share increased continuously during thirteen quarters, while until 1990 Q2 the consecutive increase lasted eight quarters. Moreover, the average growth during five consecutive quarters was 1.9 per cent and it has never been observed such an accumulated market share gain in a time span of five quarters. Considering the usual exports modeling strategy, the main (and only) candidate to explain this recent behavior of exports is the real effective exchange rate. However, considering its recent evolution, the price com- 6 For more details see, for example, Francois et al. (2007). 7 For example, Cabral and Esteves (2006) show that the significant Portuguese market share losses recorded in 2004 and 2005 occurred in sectors where it has been observed market share gains for some developing countries, namely China. One should note that the main conclusions drawn from the empirical results do not change when one controls for such a period. 10

13 Figure 1: Portuguese exports market share and real effective exchange rate. 11

14 petitiveness indicator is not able to explain these striking gains. Since the beginning of 2011, the real depreciation reached only 3.5 per cent, and this depreciation was particularly noticeable during In this paper, we suggest that the evolution of domestic demand plays an important role in explaining the recent exports market share gain. Since the first quarter of 2011, domestic demand recorded six consecutive negative quarter-on-quarter growth rates, attaining a cumulative fall close to 12 per cent. This has never happened before in the Portuguese economy and could contribute to overcome the failure of the traditional approach to model exports. Moreover, we investigate the non-linearity of this effect on exports performance. 3.3 Modeling exports behavior for Portugal Firstly, it is worth mentioning that it is particularly difficult to estimate an exports function for Portugal. In fact, following the traditional approach where exports market share is explained only by the real exchange rate leads frequently to non-reliable results. The same problem occurs in other countries. For example, Fagan et al. (2001, 2005), when developing the well-known area wide model at the ECB, included a deterministic trend to assure a long run relationship between the exports market share and the real exchange rate. In this respect, di Mauro and Forster (2008) also mention the statistical significance of a trend to explain the euro area exports performance since In particular, a negative trend was found to be stronger in the more recent period which could be related with the global integration of China. It should be mentioned that the lack of explanatory power of price competitiveness indicators occurs even when first differences 8 Furthermore, the recent evolution of price competitiveness indicators could be biased given the public wages cuts. These cuts tend to favor the usual price competitiveness indicators that are computed for the overall economy. Hence, it would be better to account only for the private sector, which, however, is not done given the lack of timely, reliable and coherent information across countries. 12

15 models are considered (i.e. disregarding the long-run relationship). This highlights the importance of other factors than price competitiveness to understand not only the long-run but also the short-run dynamics of exports. In fact, European Commission (2010) presents a panel regression for the euro area countries where the external demand and the real exchange rate are able to explain around half of the export variance. Dieppe et al. (2012) also stress that price competitiveness indicators are not able to fully explain the differences in terms of export performance across euro area countries. Concerning the Portuguese case, since 1985 and during several consecutive years there was simultaneously a strong increase of the exports market share and a remarkable appreciation of the Portuguese currency (see Figure 1). Thus, given this positive correlation, a simple regression to estimate a long-run relation between the two variables leads to a real exchange rate coefficient with the "wrong" economic sign. In this respect, at that time, several Portuguese authors presented some evidence linking the Portuguese real appreciation trend with real adjustments in the Portuguese economy (see, for example, Rebelo (1993), Cunha and Machado (1993), Esteves (1993), Gaspar and Pinheiro (1994), Pereira and Gaspar (1999)). Thus, when one considers a long time span, the real exchange rate might not be an adequate indicator to evaluate export competitiveness, as it may reflect some structural changes in the economy. In the Portuguese case, this problem can be tackled by including a deterministic variable that captures all the other factors explaining exports behavior and at the same time allowing to get the right sign for the real exchange rate. Given this ad-hoc procedure, the results obtained for the long-run solution for exports should be carefully interpreted. Typically, this additional variable is called as a non-price competitiveness indicator (see, for example, Fagan et al. (2001, 2005)). However, this type of deterministic trend does not allow for understanding and measuring the several effects underlying a simultaneous real appreciation and an increase of exports market 13

16 share. In light of the above discussion, let us first consider for the long-run relationship the case where the exports market share depends on the real exchange rate plus a log trend. 9 Concerning the short-run dynamics, besides the usual determinants of exports, we also consider domestic demand behavior while allowing for a maximum lag of four quarters for each variable. 10 As usual, the estimated models are checked by a battery of diagnostic tests. All variables are measured in logarithmic terms and a one step approach is pursued for the estimation of the error correction model. The estimation of the ECM in a single step has several advantages over the two stage procedure where the residual from the equilibrium regression is used to estimate the ECM. For instance, with the two stage approach any mistake introduced in the first step is carried forward in the second step (see, for example, Banerjee et al. (1986)). Moreover, an unrestricted ECM can be at least as efficient as a two step procedure in defining long run relationships and short run dynamics. The resulting estimated model for the whole sample period, running from the beginning of the 80 s up to the second quarter of 2012 is given by 9 A log trend seems to be a relatively suitable choice for the Portuguese case. In fact, looking at the evolution of the Portuguese exports market share, the effect of this nonprice competitiveness indicator seems to be particularly more relevant in the firstpartof the sample. 10 One should mention that the domestic demand variable did not prove to be significant in the long run relationship in the ECM models. In fact, from a theoretical point of view, it is also not clear the way the effects of domestic demand pressure operate on the longterm export performance (see, for example, Renton and Duffy (1970)). On the one hand, periods of high domestic demand pressure may stimulate investment allowing for a higher trend growth rate of exports. On the other hand, the absence of periods of very low pressure may lead to a general neglect of export opportunities. 14

17 = (3 72) (4 21) ( 1 91) ( 2 19) ( 2 28) ln (1) ( 3 38) (1 42) b = =0 269 (6 118) = 7 234[0 000] where denotes exports, corresponds to the foreign demand, is the real effective exchange rate, denotes domestic demand (including private and public consumption and investment) and is a linear trend. The HACSE t-ratios for the estimated coefficients are presented between brackets. 11 It is also reported the standard error, b, the goodness of fit statistic 2 and the usual statistic along the corresponding -value between square brackets. Concerning the long run relationship, in line with Laxton et al. (1998) and Fagan et al. (2001, 2005), an elasticity of one was imposed for the external demand coefficient which is not rejected by the data. It is also worth mentioning that the coefficient of the error correction term (i.e. the coefficient of the exports level lagged one period) is small, which denotes some persistence concerning the evolution of exports towards its long-run path. Regarding the short-run dynamics, the results do not reject a coefficient of one concerning the contemporaneous evolution of external demand, which allows to model directly exports market share behavior (see also Laxton et al. (1998) and Fagan et al. (2001, 2005)). Moreover, the results point towards an importance of lagged effects of external demand and a strong negative effect of lagged domestic demand changes. In addition, the overall goodness of fit statistics are also in line with those obtained in previous 11 Although the coefficient associated with ln is not statistically different from zero at the usual significance level, the exclusion of such deterministic variable would result in anon-significant error correction term. 15

18 empirical work on modeling exports market share (for example, Fagan et al. (2001) report a 2 around 0 2 for the euro area). In addition, we also allow for an asymmetric impact of domestic demand evolution on exports performance. In particular, we split domestic demand in two different variables, depending of its change being positive ( + ) or negative ( ), that is 12 = ½ + if 0 if 0 Allowing for an asymmetric impact, the resulting estimated model is the following (2) = (2 82) (4 69) ( 2 96) ( 2 78) ( 2 12) ln (3) ( 2 27) (0 87) b = =0 282 (6 118) = 7 707[0 000] Overall, the results are similar to those obtained in (1), with the exception of the asymmetric effects of domestic demand changes on exports behavior which cannot be disregarded. When domestic demand is falling the effects are strong and statistically significant. In contrast, when domestic demand growth is positive, it has also a negative impact on exports behavior but it is, however, not statistically significant. 13 This means that, when domestic demand falls one observes, on average, an exports performance improvement in the short-run while when domestic demand increases a negative impact is recorded, although clearly more limited in absolute terms. 12 One should mention that domestic demand change is negative in one fourth of the total number of observations. 13 The most relevant term would be the + 2 with a coefficient of 0 18 and a t-ratio of

19 Despite the strong evidence regarding the significance of the domestic demand variable, one should note that the above results should be interpreted with caution. In particular, in both models (1) and (3) one can argue that the evidence of cointegration is at most very weak as denoted by the low speed adjustment to the long run equilibrium. Within the single step framework, testing for cointegration can be performed through the significance testing of the error correction term (see, for example, Banerjee et al. (1998) and Ericsson and Mackinnon (2002)). In practice, this can be accomplished through the t-ratio of the error correction term coefficient which is called the ECM statistic. This t-ratio is used to test the null hypothesis of no cointegration (i.e., a zero coefficient for the error correction term) and the critical values can be found in Banerjee et al. (1998). As expected, based on this cointegration test, one would not reject the null hypothesis of no cointegration at the usual significance levels. As extensively discussed above, the second half of the 80 s seems to present a distinct nature. Hence, to avoid the modeling dificulties mentioned earlier, we consider the sample period starting only at the beginning of the 90 s. The estimated ECM model is the following = (3 84) (3 48) ( 2 14) ( 2 69) ( 3 19) (4) ( 4 00) b = =0 224 (5 84) = 4 847[0 001] while allowing for an asymmetric impact one obtains 17

20 = (3 36) (3 50) ( 3 25) ( 3 12) ( 3 12) (5) ( 3 14) b = =0 240 (5 84) = 5 317[0 000] Drawing on the estimated models (4) and (5) one should highlight the following. As expected, when one disregards the 80 s, the ln can be discarded without affecting the significance of the error correction term and the sign of the real effective exchange rate coefficient. Moreover, the error correction term is now much more statistically significant and one is able to reject the null of no cointegration with a significance level of 10 per cent and almost with a significance level of 5 per cent (the asymptotic critical values are 2 89 and 3 19, respectively). In addition to this evidence in favour of cointegration, one should also mention that the coefficient of the error correction term, that is, the speed of adjustment to the long-run equilibrium, is the same found by Fagan et al. (2001, 2005) for the euro area. Concerning domestic demand, such variable appears again quite significant in both models and the finding that negative changes in domestic demand are clearly more relevant that positive ones holds. In particular, regarding the market share gains observed since the beginning of 2011 (corresponding to an accumulated gain of almost 10 per cent), model (5) would have predicted a market share gain above 11 per cent, conditional on the observed evolution for the independent variables over that period. Such prediction is almost entirely explained by domestic demand developments which highlights its importance in the latest episode of exports market share gain observed in the Portuguese economy. 18